At Lotus F1 Team, Thomas Mayer has the job of overseeing work to build "a plane that drives on the road".
Building a car capable of winning the Formula One championship is more science than art. To identify the most aerodynamic design, the racing team tests the performance of more than 500 models each week, crunching data using its 38 teraflop computing cluster.
Lotus constantly collects data to refine these computer models — ranging from how air flows over prototype designs in its wind tunnel to telemetry from its cars as they race around tracks, with each vehicle relaying data every 100th of a second from more than 150 on-board sensors.
"The business is highly competitive — it's not only the speed of the car, but also how fast we can develop, how fast we can iterate from the beginning to the end of the season," said Mayer, who is chief operating officer of the racing team.
The trove of data Lotus collects from its cars during a race also shapes the strategy the pit crew and the driver take during that event. These team relies on some 36 servers to run Lotus's track-side applications: its track, performance and race strategy systems — all of which have been developed in-house by Lotus.
The need to have these systems at hand during a race means that when the Lotus teams jet off to one of the 19 races they compete in worldwide, they take their servers with them.
"We have to physically transport them around the world. It's really expensive and we have problems with damage, power supply and other operational problems," Mayer said at the Cloud World Forum event in London yesterday.
Within three years, Lotus wants to host these jet-setting systems in Microsoft's Azure infrastructure instead, targeting datacentres close to the tracks where the events take place.
"It would be much nicer to have a big cable and plug into the cloud," he said.
The main barrier to doing so, Mayer said, is latency.
"With our performance system, we want to gain something like a two-second window ahead of what we are forecasting is going to happen, so we have something like two seconds to make our decisions," he said.
"We are speaking with network providers about how to overcome this but if you have a dedicated fibre line then you have again the redundancy problem, if someone cuts this line we are dead."
Lotus' other primary concern when it comes to storing data and applications on a cloud platform is security.
"Data is very sensitive and we can lose years of development if it goes public," Mayer said.
Sensitive and commercially-valuable data is scrambled using 128-bit encryption and is also spread across multiple drives to limit the loss in the event of someone stealing one of the drives.
Lotus also plans to move a range of other systems to Azure, such as Microsoft Dynamics ERP and CRM system.
"We want to focus on making the car faster and don't want to waste time on something that's a commodity."
The company is even testing cloud infrastructure for its high-performance computing needs, piloting a system that would provide additional computing power at the beginning of the season when the team is working on building the car.